Analysis of Signaling Pathways in 90 Cancer Cell Lines by Protein

Analysis of Signaling Pathways in 90 Cancer Cell Lines by Protein Lysate Array Kanchana Natarajan Mendes,†,# Daniel Nicorici,†,‡,# David Cogdell,†,# Ioan Tabus,‡ Olli Yli-Harja,‡ Rudy Guerra,§ Stanley R. Hamilton,† and Wei Zhang*,† Department of Pathology, the University of Texas M. D. Anderson Cancer Center, Houston, Texas 77030, Institute of Signal Processing, Tampere University of Technology, Tampere, Finland, and Department of Statistics, Rice University, Houston, Texas 77251 Received March 30, 2007

Multiple signal transduction pathways play a crucial role in cancer development, progression, and response to different therapies. An important issue is whether common signal transduction pathways are ubiquitously altered in all cancer types and some unique pathways are involved in different cancer types. Another important issue is whether and how transduction signaling molecules are heterogeneously expressed and activated in different cancer cells within and between cancer cell types. Methods: To gain insight into these issues, we assembled a protein lysate array with 90 different cell lines of 12 different cell types. Each sample is diluted 2-fold six times, and samples from the dilution series were printed three times on the array. We then measured the expression levels and phosphorylation status of 52 different signaling proteins with specific antibodies and carried out statistical hierarchical clustering analysis. Results: The most significant finding based on the cluster analysis was that the cell lines did not group based on tumor types, suggesting that the signaling pathways studied were commonly activated in most of the tumor types cultured in vitro. As expected, related proteins associated with specific signaling pathways clustered together, and analysis of the 30 most differentially expressed proteins revealed the PI3-K signaling pathway was upregulated in several different tumor types and the VEGF-angiogenesis pathway was downregulated in hematopoetic cancers. Another important observation, with clinical implications was that EGFR was the most heterogeneous among all the cell lines. We also observed signaling pathways unique to specific types of cancers such as the inverse relationship between p16ink and Rb, and the EGFR mediated pathway activation characteristic of pancreatic cancers. Conclusions: Using reverse phase lysate array analysis in this study, we were able to determine potential relationships and signaling pathways, both common and unique, to different types of cancer using cell lines in vitro. This data could be utilized for mining information related to an individual cancer of interest and combined with morphological and genomic profiles would help in creating a combination of expression markers and/or functional signaling maps for specific cancer diagnosis and therapy. Keywords: Reverse-phase lysate array • Cancer • Signaling pathways • Proteomics • Clinical therapy

Introduction Neoplasia results from a multistep dysregulation of processes such as growth signaling, apoptosis, angiogenesis, replication, and cell migration.1 At molecular level, alteration of these processes involves a large number of translationally and posttranslationally modified signaling proteins. However, given the enormous individual tissue and cell type heterogeneity among * To whom correspondence should be addressed: Wei Zhang, Ph.D., Department of Pathology, Unit 85, the University of Texas M. D. Anderson Cancer Center, 1515 Holcombe Blvd., Houston, Texas 77030. Tel, 713-7451103; fax, 713-792-5549; e-mail, [email protected]. † The University of Texas M. D. Anderson Cancer Center. ‡ Tampere University of Technology. # Equal contribution as co-first authors § Rice University. 10.1021/pr070184h CCC: $37.00

 2007 American Chemical Society

tumors and even the cell lines that were derived from tumors, traditional approaches that focus on identification of individual abnormalities in tumor cells have not been sufficient in either understanding the abnormal processes involved in cancer or in the identification of robust markers for diagnosis and prognosis. We would expect better prediction of disease behavior if we could profile and classify multiple components of aberrant cell signaling pathways simultaneously.2 Consequently, the use of high-throughput techniques that map entire networks rather than individual markers would be necessary and effective. Advancements in genome-wide analysis opened up several avenues to study diseases on a global level. DNA profiling studies such as microarrays have proved to be powerful tools Journal of Proteome Research 2007, 6, 2753-2767

2753

Published on Web 06/12/2007

research articles

Natarajan Mendes et al.

Table 1. Types of Cancers, Cell Lines Used, and Their p53 Status type of cancer

WT p53

Breast

MCF10A, HBL100, T-47D, MCF-7

Colon Glioma Kidney Leukemia Lung

RKO, HCT116 +/+ U87, LN229 (WT function), SNB19-P, D54, A172 HEK293 OCI-AML3, BV173 A549, NCI-H460

Lymphoma Melanoma Ovarian Pancreatic Prostate Sarcoma

A375SM TOV-21 G, PA-1 Capan 2 LNCaP NIH-3T3, A204

mutant p53

MDA-MB-468BP5, P11MDA-MB-435, MDA-MB-231, MDA-MB-453, BT474.M1, Sk-Br-3 KM12C, KM12SM, HT29, DLD-1, SW480 U373, PL

CEM-UV, Jurkat NCI-H23, NCI-446, NCI-N417, H322, H596, NCI-H226, NCI-H226Br Raji MEWO OV-90, OVCAR-3 Capan 1, CF, E6E7, PANC-1, Miapaca2 Du145, PC-3, PC2a RD, A431, SKLMS 1

in providing sophisticated multiplex panels for classifying cancers and in therapeutic studies.3-6 However, these genomebased technologies fail to take into account the multiple protein products associated with the genes, their functional significance, post-translational modifications, subcellular and tissue distribution, and protein interactions. In addition, as most drugs target proteins, a protein-profile-based strategy might be more relevant to therapy.7 The field of proteomics has come to encompass the systematic analysis of protein populations with a goal of concurrently identifying, quantifying, and analyzing large numbers of proteins in a functional context and has enormous potential in identifying proteins associated with different disease states. In the past decade, new large-scale proteomic technologies have been introduced that may help to identify the complex signaling interactions involved in growth of cancer cells.8 One such novel technology is reverse-phase protein lysate microarray,9,10 which is a valuable technique to quantify a panel of specific proteins in cell or tissue samples. Proteins from a large number of samples are robotically spotted onto a coated glass slide. Specific antibodies are used to detect the protein level of these samples on each slide. Because of the density of spots that can be included on one slide, multiple dilutions and replicates are often incorporated into the design to increase the data robustness. Studies using this technology have shown promising results in both monitoring disease-related protein expression as well as in investigating the molecular and cellular effects of therapeutic agents.11-14 One comprehensive study using this technology in cancer biology involved the proteomic profiling of 60 different human cancer cell lines (NCI-60) using 52 different antibodies to compare the protein expression patterns with mRNA expression data obtained by cDNA and oligonucleotide arrays for those same genes.10 This study led to the discovery of two promising markers, villin and moesin, for distinguishing colon from ovarian adenocarcinomas and defined a class of cell-structure-related molecular species in which the mRNA/protein correlation was high. Studies in our lab have previously used this technology to study protein profiles in gliomas tissues and breast cancer cell lines.15-17 Our aim, in this study, was to compare various signaling networks such as the apoptotic pathway, the cell cycle pathway, and the effect of p53 mutations on these pathways between different types of cancer using available cell lines. With the use of reverse-phase lysate array technology and a cell line array comprised of 90 different cancerous lines, we demonstrated that a few common signal transduction pathways were 2754

Journal of Proteome Research • Vol. 6, No. 7, 2007

null p53

HCT116 -/U251, LNZ308

U937, THP-1, HL-60, NB4 NCI-H358, Calu-6, NCI-H12

SK-OV-3 ASPC-1 A673, SAOS 2, LM 2

ubiquitously altered in all cancer types. We also identified unique pathways that were specific to particular types of cancers.

Materials and Methods Cell Line and Protein Lysate Preparation. Cell lines lysates were collected in buffer containing 20 mM Tris, pH 7.6, 150 mM NaCl, 5 mM EDTA, and 0.5% NP40. Protein concentrations of the lysate solutions were quantified by the Bradford method (Bio-Rad Laboratories, Hercules, CA). Table 1 lists the known phenotypes and genotypes of the 90 cell lines. Lysate Array Design and Antibodies. Lysate solutions (20 µg/µL) were serially diluted 2-fold six times and printed on PVDF-coated glass slides in triplicate using a robotic spotter (G3, Genomics Solutions) described by Jiang and colleagues.16 Each slide was incubated with 53 specific antibodies, which was described in Table 2. Proper secondary antibodies (Vector Laboratories, Burlingame, CA) were added after primary incubations. Detection of Proteins on the Slides. Detection of specific proteins was examined by DakoCytomation catalyzed signal amplification system kit (CSA, DakoCytomation, Carpinteria, CA) according to the manufacturer company. Briefly, the slides were blocked with the biotin blocking kit for 5 min, and blocking continued with protein block reagent for 10 min. Primary antibodies were diluted 1:100 to ∼1:200 in antibody dilution buffer and incubated on slides at 25 °C for 2 h, and secondary biotinylated antibodies (1:4000 to ∼1:10 000) were added and incubated for 1 h. Streptavidin-biotin-peroxidase complex from the amplification kit was incubated for 15 min for signal amplification. After application of hydrogen peroxide for development of slides, the slides were dried at room temperature. β-Actin and blank without any primary antibody were used as positive and negative controls, respectively, in each set of experiments. Signals were scanned at 1200 dpi resolution, and spot images were converted to a 16-bit grayscale. The spots were analyzed by ArrayVision (Imaging Research, Inc., Catharines, Ontario, Canada) for quantification. Evaluation of Lysate Array Data. Three 2-fold serial dilutions were printed for each sample, yielding 18 data points per each sample. The lysate array contained 90 cell line samples (52 antibodies where β-actin is not included). The protein expressions were quantified using the nonlinear modeling of the protein method described in Tabus et al.18 The protein expression levels were normalized against β-actin protein from each

research articles

Signaling Pathways in Cancer Table 2. Antibodies Applied in Lysate Array to Detect Proteins pathway

PI3-K

name of antibody

company

Akt, Akt[pThr308], Akt[pSer473], Bad[pSer136], mTOR, mTOR[pSer2448], PTEN, RSK, RSK[pThr573], PDK, PDK[pSer241] Bad PI3-K

Growth factor/Integrin signaling

MAPK, PDGFR β, PDGFR β [pTyr751], PAK, PAK[pSer199/204], Zap70, Zap70[pTyr493] CD11b, Src, Src[pTyr418], Src[pTyr529] c-abl, c-kit, IGFBP2, IGFBP3, IGFBP5, Tie-2 EGFR

Apoptosis

cleaved caspase 8, cleaved caspase 9 Bax, Bcl-2, Cytochrome b Cytochrome c, Cytochrome IV-2, Cytochrome IV-Vb

Cell cycle

AIF Cdk4, Cdk7, Cyclin D3, p53 (Do-1) p16, p19, RB c-myc p53 1801

Others

CD154, HXK1, ND2

Normalization control

β-actin

production lot, and then, median quantile normalization method was applied for normalizing the between arrays variability.19 Protein features with the highest discrimination power were selected based on the BSS/WSS score for different statuses.20 We chose the subset of the most differentially expressed proteins based on p-values and/or false-discovery rate (FDR) which is the expected fraction of false positives in the list of proteins declared significant.21,22 The p-values are computed by randomizing the labels of the cell lines for an ensemble of 1000 runs using the BSS/WSS score. An extension to FDR is q-value that was introduced in Storey et al.23 The p-value is stricter than the q-value. The q-value is defined as the smallest FDR at which a particular protein is still on the list of the proteins declared significant.23 The hierarchical clustering with average linkage and correlation coefficient as metric is used for presenting the results. The hierarchical trees with p-values from Figure 1C,E were generated using the R package PVCLUST, which gave the AU (Approximately Unbiased) p-values (red values) and BP (Bootstrap Probability) values (green values) in percent. Clusters with AU larger than 95% are highlighted by rectangles, which are strongly supported by data.24

Results Hierarchial Clustering Analysis of the Protein Profiles of all Cell Lines Used Shows a Strong Relationship among Different Signaling Pathways and Confirms the Heterogeneity between Individual Cell Lines. We analyzed the expression and phosphorylation of 52 different proteins that have been shown to be involved in aberrant signaling pathways associated with tumor development, including apoptosis, cell cycle regulation,

Cell Signaling Technology, Inc., Beverly, MA 01915 Santa Cruz Biotechnology, Inc., Santa Cruz, CA 95060 BD Biosciences Immunocytometry Systems, San Jose, CA 95131 Cell Signaling Technology, Inc., Beverly, MA 01915 BioSource International, Inc. (Invitrogen) Camarillo, CA 93012 Santa Cruz Biotechnology, Inc., Santa Cruz, CA 95060 Zymed Laboratories, Inc. (Invitrogen) Carlsbad, CA 92008 Cell Signaling Technology, Inc., Beverly, MA 01915 Santa Cruz Biotechnology, Inc., Santa Cruz, CA 95060 Molecular Probes, Inc. (Invitrogen), Eugene, OR 97402 eBioscience, Inc. San Diego, CA 92121 Santa Cruz Biotechnology, Inc., Santa Cruz, CA 95060 BD Biosciences Immunocytometry Systems, San Jose, CA 95131 GeneTex, Inc. San Antonio, TX 78245 Oncogene Science, Inc., Uniondale, NY 11553 Santa Cruz Biotechnology, Inc., Santa Cruz, CA 95060 Sigma Chemical Co. St.Louis, MO 63178

and angiogenesis in 90 cancer cell lines of 12 different cancer types using the lysate array. Antibody specificity was assessed and confirmed by Western blotting prior to use in the lysate array as shown in Jiang et al.16 All the raw data can be accessed on our Web site at www.mdanderson.org/∼genomics. Table 1 summarizes the cell lines used and the type of cancer they are associated with. Figure 1A shows the layout of the protein lysate array slide and a few examples of the antibodies used for staining. An overall clustering analysis revealed that there was a close relationship among Src, Phospho-Src, PDGFR, Akt, mTOR, and PI3-K, all of which were signaling molecules associated with the phosphoinositide 3 kinase pathway (Figure 1B). Similarly, cell cycle related proteins such as Cdk4, Cyclin D3, and Cdk7 were clustered together as were apoptosis related proteins like Bcl2, cytochromes, caspases 8 and 9, and phosphBad. In addition, all the insulin-like growth factor binding proteins (IGFBPs), IGFBP2, IGFBP3, and IGFBP5, also were clustered together confirming modulation of entire pathways associated all the cancer cell lines. A cluster dendrogram analysis of the protein profiles revealed that, of the various proteins profiled, the most variable protein was the epidermal growth factor receptor (EGFR), as it did not cluster with any of the other statistically significant clusters of proteins profiled (Figure 1A,C). The variability of relative EGFR expression in all the cell lines used is shown in Figure 1D. Proteins such as IGFBP3 and hexokinase1 (HXK1), both of which are involved in glucose metabolism pathways, also clustered together along with c-kit, which is an interesting observation since imatinib, a small molecule c-kit inhibitor, has been used as therapy in gastrointestinal tumors based on decreased glucose uptake by the tumor cells.25 Journal of Proteome Research • Vol. 6, No. 7, 2007 2755

research articles

Figure 1. (Continued) 2756



Signaling Pathways in Cancer

research articles

Figure 1. (Continued) Journal of Proteome Research • Vol. 6, No. 7, 2007 2757

research articles


Figure 1. (A) Protein lysate array design (using a Colloidal Gold stained slide for total protein that has been overlaid with the subtype grouping reflecting lysate source) and examples of applied antibodies on slides. To show several array images of the 52 antibodies used, we used β-actin antibody on the slides and used them as controls, and we selected slides containing several important signaling pathway proteins, including Akt (pSer308), and EGFR. (B) Hierarchical clustering of the levels of expression of 52 proteins in the 90 cancer cell lines. Red and green colors indicate high and low protein expression, respectively. (C) Hierarchical clustering of 52 proteins. Values at branches are AU p-values (left), BP values (right), and cluster labels (bottom). Clusters with AU g 95 are indicated by the rectangles. (D) Relative expression levels of EGFR in the cell lines used in the lysate array analysis. (E) Hierarchical clustering of 90 cell lines used in the lysate array. Values at branches are AU p-values (left), BP values (right), and cluster labels (bottom). Clusters with AU g 95 are indicated by the rectangles.

Clustering analysis was also done to determine the relationship between each of the cell lines within different tumor types using all the proteins. This analysis showed that the cell lines were clustered heterogeneously and cell lines did not group based on tumor types, suggesting that these signaling pathways are commonly activated in most of the tumor types cultured in vitro (Figure 1E,B). The p53 Mutational Status in the Cell Lines Is Closely Related to the PI3-Kinase Pathway. We determined the p53 status of each of the 90 cell lines used by reviewing the literature26 and evaluated the effect of the p53 mutational status on the protein profile generated with the lysate array. The p53 status of each of the cell lines is listed in Table 1. Interestingly, of the 5 most significantly differentially expressed (Q e 0.15) proteins depending on the wild-type p53 status of the cells, 4 were part of the platelet derived growth factor (PDGFR) pathway including PDGFR, Src, and phospho-Akt, and they 2758


were all upregulated compared to cells that had either a mutant or a null-p53 genotype (Figure 2A and Supporting Information Table S1). Upon analyzing a larger number (30) of protein profiles dependent on the p53 status of cells, we observed that wildtype p53 containing cells had higher expression levels of proteins associated with the PI3-K pathway as compared to cells that were either mutant or null for the p53 gene (Figure 2B and Supporting Information Table S2). We also did not observe any correlation of the p53 status in the cells with its downstream effector genes such as p21WAF1 and Bax. Surprisingly, IGFBP3, which is also a downstream target of p53, had higher expression levels in p53 null cells along with IGFBP2. Hematopoietic Cancers Have Lower Expression Levels of Proteins Involved in Angiogeneis. To determine the differences between solid and non-solid tumors, we divided the 90 cell lines into two groups: leukemia and all the rest. We then did a


research articles

Figure 2. Clustered image map relating (A) the 5 most significant differential protein expression with p53 wild-type status in the 90 cancer cell lines. (B) Protein expression of 30 proteins with p53 status in the 90 cancer cell lines. Journal of Proteome Research • Vol. 6, No. 7, 2007 2759

research articles


Figure 3. Clustered image map relating protein expression in non-solid tumors (leukemia) compared to the rest of the 82 cell lines.

hierarchical clustering analysis based on this grouping. The angiopoeitin receptor Tie2 showed the most significant difference in expression between the two groups (Figure 3 and Supporting Information Table S3). In addition, Tie2 was closely clustered with cytochrome IV-2, cytochrome-c, and cleaved caspase-8, suggesting that the angiogeneis-related VEGF pathway was low in non-solid tumors as expected. In addition, the CD40 ligand, CD154, which is critical for cell-mediated and humoral immunity, was downregulated in these non-solid tumors. Analysis of Individual Cancer Types. Differential expression patterns and protein profiles specific to each type of cancer was then analyzed by grouping all the cell lines associated with each type of cancer and comparing them to the rest of the cell lines in the lysate array. 1. Gliomas. The protein profile of gliomas was compared to the profiles of all the other cell lines grouped together. Consistent with the invasive nature of gliomas, we found that proteins associated with the PI3-K pathway were differentially upregulated compared to the other cell lines (Figure 4A and Supporting Information Table S4). One of the most significant differentially expressed protein was the Cyclin-dependent kinase (CDK) 4, a master integrator that couples mitogenic and antimitogenic extracellular signals with the cell cycle. 2. Colon Cancer and Sarcoma. Comparison of the colon cancer cell lines to all the other cell lines grouped together revealed that these cancers have high levels of retinoblastoma (Rb), and the proto-oncogene c-myc proteins compared to the other types of cancer (Figure 4B and Supporting Information Table S5). Similar grouping and analysis done comparing the 2760


sarcoma cell lines to the rest revealed significantly lower levels of Rb and RSK in these cell lines along with a high differential expression of the CDK4 inhibitor p16ink (Figure 4C and Supporting Information Table S6). This inverse relationship between Rb and p16ink was specific mainly to sarcomas and is illustrated in Figure 4D. 3. Lung, Prostate, and Pancreatic Cancer. Cluster analysis of the lung cancer cell lines showed a lower expression of MAPK and cytochrome IV-2 compared to the rest of the cancer types. Consistent with effects on respiration, they also expressed much higher levels of total and phosphorylated pyruvate dehydrogenase kinase (PDK) (Figure 5A and Supporting Information Table S7). We also saw increased expression levels of IGFBP2 in both pancreatic and prostate cancer, when they were compared with the rest in two separate analyses (Figures 5A,B, Supporting Information Tables S7 and S8). Pancreatic cancer was the only type of cancer in our analysis where the EGFR was significantly differentially expressed and clustered with another protein, phosphorylated c-Src (Figure 5C and Supporting Information Table S9). In addition, the tyrosine kinase c-Abl was expressed less in these cell lines than other cancer types.

Discussion Using the direct approach of studying the proteomic circuitry by the reverse-phase lysate array analysis in this study, we were able to determine potential relationships and signaling pathways, both common and unique, to different types of cancer using cell lines in vitro. On the basis of the proteomic profiling


research articles

Figure 4. (Continued Journal of Proteome Research • Vol. 6, No. 7, 2007 2761

research articles


Figure 4. Clustered image map relating protein expression in (A) gliomas compared to the rest of the 81 cell lines, (B) colon cancers compared to the rest of the 82 cell lines, (C) sarcomas compared to the rest of the 76 cell lines, and (D) comparative relative p16 (INK4a) and Rb expression levels in all the cell lines analyzed.(Continued) 2762



research articles

Figure 5. (Continued) Journal of Proteome Research • Vol. 6, No. 7, 2007 2763

research articles


Figure 5. Clustered image map relating protein expression in (A) lung cancers compared to the rest of the 77 cell lines, (B) prostate cancers compared to the rest of the 85 cell lines, and (C) pancreatic cancers compared to the rest of the 82 cell lines.

of 90 different cell lines associated with 12 types of tumor by 52 different antibodies, we observed several interesting correlations. The most significant finding was that the cell lines did not group based on tumor types, suggesting that the signaling pathways studied were commonly activated in most of the tumor types cultured in vitro. As expected related proteins associated with specific signaling pathways clustered together, the PI3-K signaling pathway was upregulated in several different tumor types and the VEGF-angiogenesis pathway was downregulated in hematopoetic cancers. Another important observation, with clinical implications, was that EGFR was the most heterogeneous among all the cell lines. Surprisingly however, we did not see a significant correlation between p53 and its downstream effector genes such as p21 and Bax. A unique aspect, specific only to sarcomas, was the inverse relationship between p16ink and Rb, while the EGFR mediated tyrosine kinase activation and signaling pathway was characteristic of pancreatic cancers. The data generated in this study is available on our Web site and could be utilized for mining information related to an individual cancer of interest. This data, combined with morphological and genomic profiles would help in creating a combination of specific cancer type expression markers and/or functional signaling maps for preclinical studies. The epidermal growth factor receptor (EGFR) is a cellular transmembrane receptor with tyrosine kinase enzymatic activity and is involved in cancer cell proliferation, apoptosis, angiogenesis, invasion, and metastasis.27,28 Understanding and targeting the epidermal growth factor receptor (EGFR) to 2764


abrogate EGF-mediated signaling pathways in malignancy has been used as a therapeutic strategy. Several anti-EGFR drugs are in Phase III clinical development as single agent or in combination with other anticancer modalities. Examples include Cetuximab (Erbitux), a chimeric human-mouse monoclonal immunoglobin (Ig) G1 antibody, which blocks ligand binding and functional activation of the EGFR for the treatment of advanced colorectal cancer, and Gefitinib (Iressa), a small molecule EGFR-selective inhibitor of tyrosine kinase activity which blocks EGF autophosphorylation and activation for the treatment of chemoresistant non-small cell lung cancer patients.29 However, some issues related to side effects and nonresponders do remain in some of these clinical trials,30 which may in part be due to the heterogeneous expression of EGFR, as we observed in our array data, suggesting that individualized patient responses to EGFR-directed therapy might be highly varied, making it necessary to determine EGFR levels in patients prior to treatment. Several pathways mediated by tyrosine kinases including EGFR and Src are known to be either overexpressed, or constitutively activated, in pancreatic cancer due to transactivation by gastrointestinal peptides.31 This is consistent with the data that we obtained from our protein lysate array where we saw an increased expression of both Src and its activated form, phospho-Src 418, in pancreatic cancer cell lines compared to the rest of the cell lines. Hence, blocking receptor tyrosine kinases and nonreceptor, cytoplasmic tyrosine kinases represents a rational approach to treat pancreatic cancer. Cetuximab and erlotinib, the monoclonal antibodies against


EGFR-1 (ErbB-1), showed promising activity in Phase II and Phase III trials, and their combination with the cytotoxic agent gemcitabine resulted in synergistic antitumor activity,32 potentially due to the suppression of the receptor tyrosine kinasemediated pathways downstream of EGFR. In addition, the lower levels of c-abl that we observed in our studies might correlate with the highly invasive properties and the aggressive nature of pancreatic cancers. In normal cells, c-Abl phosphorylates the cytoskeleton-associated adaptor protein, Crk, causing disassociation of Crk from the Crk-associated substrate (CAS) and disassembly of Crk/CAS complexes. c-Abl-induced disruption of Crk/CAS complexes inhibits cell migration and promotes apoptosis in normal cells, implying that targeting pathways associated with c-Abl-mediated cell migration may inhibit the invasiveness of pancreatic cancers.33 Another tyrosine kinase, the platelet-derived growth factor receptor (PDGFR), is also implicated in the development and progression of different tumors in numerous studies34,35 and uses the Ras/Raf/MEK/ERK and the Ras/PI3-K/PTEN/Akt signaling cascades to transmit signals from their receptors to regulate growth and tumorigenesis. The studies done by McCubrey et al. showed that restoration of wild-type p53 in p53 deficient prostate cancer cells results in their enhanced sensitivity to chemotherapeutic drugs by modulating molecules associated with the cell cycle and increased expression of pathways downstream of the PDGFR.36 Both the PDGFR and its activated phosphorylated form were closely related to the p53 status of the cells in our lysate array data confirming the previous findings. There are several lines of evidence implying a relationship between the p53 status of cells and the activation of the PI3-K pathways, as was observed in our lysate array data, where wildtype p53 cells (usually lower levels of p53 expression) correlated with increased differential expression of proteins and their modified forms involved in the PI3-K pathway such as PI3-K, Akt, mTOR, Src, and MAPK. Previous literature on wild-type p53 containing human ovarian carcinoma cells indicates that the PI3-K/AKT signal transduction pathway mediates p21 expression and suggests that this pathway contributes to cell cycle regulation promoted by p53 in response to drug-induced stress.37 Recent work done by Cho et al. has shown that inhibition of the PI3-K/Akt pathway downregulates the expression of the human cervical cancer oncogene HCCR-1, which in turn acts as a negative regulator of p53.38 Other studies using breast cancer and T-cell acute lymphoblastic leukemia cells have described chemoresistance acquired by inhibiting the p53 pathway through an mTOR-dependent PI3-K-Akt/PKB pathway.39 In this light, a feasible therapeutic strategy would be to activate the p53 pathway by suppressing the PI3-K pathway. AMP-activated protein kinase (AMPK) is a member of a metabolite-sensing protein kinase family.40 Activation of AMPK inhibits proliferation of various cancer cell lines in vitro and in vivo by increasing p21CIP, p27KIP, and p53.41 Thus therapeutically, the constitutive activation of the PI3-K-Akt-mTOR signaling cascade reported in many cancers including glioblastoma, melanoma, and advanced prostate cancer can be suppressed by AMPK since AMPK inhibits mTOR signaling downstream of Akt, and inhibition of mTOR pathway has been reported to inhibit tumor growth and metastasis in experimental animal models as well as in cultured cells.42 In addition to AMPK, Mdm2, an important mediator of growth and survival signaling in the PI3-K/Akt pathway, may be a potential molecule that may serve as an mediator between the two pathways as the

research articles major contribution of Mdm2 to the development of cancer is through a tight inhibition of the activities and stability of the tumor suppressor p53.43 This implies that p53 status and signaling networks that modulate p53 levels and activity are perhaps important determinants in combination therapeutic strategies while targeting the various downstream mediators of the PI3-K pathway. Various angiogenic factors, such as vascular endothelial growth factor (VEGF) and angiopoietins (Angs), are thought to be associated with leukemia cell growth. The Tie2 receptor, upon binding to Angs, are known to inhibit apoptosis by activating PI3-K.44 This finding is in line with our data where we observe an increase in the downstream modulators of the PI3-K pathway, Src, mTOR, and phospho-mTOR. Our lysate array data, along with other studies, indicates that in human leukemias there are low levels of caspase 8 and cytochrome-c activation and high levels of Bcl2, suggesting that mitochondrial apoptosis pathways are downregulated providing a rationale for designing therapeutic strategies based on activating these apoptosis pathways.45,46 Our lysate array analysis also showed lower levels of Tie2 in leukemia cells when compared to other types of cancer, consistent with the fact that non-solid tumors such as some types of leukemia would require less of angiogenesis-specific proteins compared to solid tumors. However, recent studies have indicated the importance of angiogenesis in the pathophysiology of several hematologic malignancies, and this has led to the use of anti-angiogenic therapy in pre-clinical trials of some types of leukemia such as CML and CLL, although none of these therapies target or involve Tie-2.47 Another molecule that we found to be downregulated in leukemias in our analysis was the CD40 ligand CD154 (CD40L). Farahani et al. have reported that there is an interplay between the CD154 and the VEGF pathway in the survival of leukemia cells. In their experiments, these investigators demonstrated neutralization of the VEGF pathway, with either an anti-VEGF antibody or receptor tyrosine kinase (RTK) inhibitor, leading to a considerable reduction in CD154 enhanced apoptotic resistance in the cells.48 Clinical trials using autologous CD154adenovirally transduced leukemia cells as a cellular vaccine were capable of inducing a cellular anti-leukemia immunity, and had a direct effect on leukemia cells by inducing Fas (CD95)-dependent leukemia-cell apoptosis.48,49 Both the oncogene c-myc and the tumor suppressor Rb have been reported in colorectal cancers50 and were the most significantly correlated proteins and highly differentially expressed in our lysate array data when we compared colon cancer cell lines to all the rest of the cell lines used. With this is mind, studies conducted on colon carcinoma cell lines using phosphorothioate modified antisense oligonucleotides (ASO), complementary to the c-myc translation initiation site, showed specific growth inhibition and increased differentiation and correlated with a decrease in both c-myc and Rb levels, suggesting a signaling crosstalk between the two proteins.51 Another interesting observation that we came across while analyzing sarcoma lysate data was that this was the only group in which there was an inverse relationship between p16ink and RB. In most cases, deregulation of E2F transcriptional activity as a result of alterations in the p16 (INK4a)-cyclin D1-Rb pathway is a hallmark of human cancer.52 These results indicate that the inverse relationship between p16ink and Rb in the cell cycle signaling networks could be useful in prognostically predicting sarcomas. Journal of Proteome Research • Vol. 6, No. 7, 2007 2765

research articles The insulin-like growth factor (IGF) axis is a complex system composed of 2 mitogenic ligands, IGF-I and -II, 2 receptors, IGF-1R and IGF-2R, and 6 binding proteins, IGFBP-1 to -6. The IGFBPs exert their actions through their regulation of IGF bioavailability for IGF receptors. IGFBP-2 is a major IGFBP in the prostate and in seminal fluid, and its levels, which are elevated in many malignancies, are markedly increased in prostate cancer correlating with the aggressiveness of the disease.53 Thus, it was not surprising to find elevated differential expression of IGFBP2 in the prostate cancer samples compared to the other cancers in our data as well. Hence, targeting IGFBP2 and pathways downstream of it may serve as a basis for designing therapy strategies in pancreatic cancers. Cancer progression is characterized by the accumulation and cooperation of multiple genetic mutations and is orchestrated via protein alterations along multiple pathways that drive proliferation, block differentiation, or inhibit apoptosis. Key regulatory proteins such as protein kinases and phosphatases mediate cellular signaling circuitry, and their aberrant function is frequently central to the pathogenesis of cancer,54 leading to heterogeneity associated with most cancer types. Entire pathway information obtained from lysate arrays, such as what we have presented in this paper, could be used to select multiple drug targets within a signaling cascade rather than a single drug target, thereby potentially allowing for less drug toxicity.55,56

Concluding Remarks As seen in all the reports discussed above, signal transduction in cancer cells is a sophisticated process that involves receptor tyrosine kinases (RTKs) that eventually trigger multiple cytoplasmic kinases. Several key cellular signaling pathways that work independently, in parallel, and/or through interconnections have been identified that promote cancer development. Three major signaling pathways that have been identified as playing important roles in cancer include the phosphatidyl inositol-3-kinase (PI3-K)/AKT, protein kinase C (PKC) family, and mitogen-activated protein kinase (MAPK)/Ras signaling cascades.57 In our lysate array, we found the PI3-K pathway to be the most differentially expressed and correlated among the 12 types of cancers that we studied. In clinical trials, highly selective or specific blocking of only one of the kinases involved in these signaling pathways has been associated with limited or sporadic responses. Thus, in addition to gaining knowledge about potential markers associated with a particular type of cancer and common pathways and crosstalk associated with a variety of cancers in general, the protein lysate array technology that we have used would be useful in designing therapy strategies. Analysis of different signaling pathways would hence improve our understanding of the complexity of signal transduction processes and their roles in cancer allowing simultaneous inhibition of several key kinases at the level of receptors and/or downstream kinases, thereby helping to optimize the overall therapeutic benefit associated with anticancer agent.

Acknowledgment. We thank Ellen Taylor and Yu Jack Jia for their technical assistance. We thank Drs. Dihua Yu, Jonathan Kurie, Jinsong Liu, Paul Chiao, Jonathan Trent, Steve Kornblau, and Renata Pasquilini for providing cell lines and protein extracts. The Cancer Genomics Core Lab is supported by the Tobacco Settlement Fund to M. D. Anderson Cancer Center as appropriated by the Texas Legislature, by grants from the Michael and Betty Kadoorie Foundation and the Goodwin Fund, and by an NCI Cancer Center Support Grant CA-16672. 2766



Supporting Information Available: The following supplementary data tables are available: (S1) The 5 most differentially expressed proteins in cell lines with wild-type p53 compared to other p53 status, (S2) the 30 most differentally expressed proteins in cell lines with wild-type p53 compared to either mutant or null p53 status, (S3) the 20 most differentially expressed proteins in leukemia cell lines compared to all other cancer types, (S4) the 30 most differentially expressed proteins in glioma cell lines compared to all other cancer types, (S5) the 10 most differentially expressed proteins in colon cancer cell lines compared to all other cancer types, (S6) the 15 most differentially expressed proteins in sarcoma cell lines compared to all other cancer types, and (S7) the 5 most differentially expressed proteins in lung, (S8) prostate, and (S9) pancreatic cancer cell lines compared to all other cancer types. This material is available free of charge via the Internet at http://pubs.acs.org. References (1) Hanahan, D.; Weinberg, R. A. The hallmarks of cancer. Cell 2000, 100 (1), 57-70. (2) Chatterjee, S. K.; Zetter, B. R. Cancer biomarkers: knowing the present and predicting the future. Future Oncol. 2005, 1 (1), 3750. (3) Bagg, A. Molecular diagnosis in lymphoma. Curr. Hematol. Rep. 2005, 4 (4), 313-23. (4) Modlich, O.; Prisack, H. B.; Bojar, H. Breast cancer expression profiling: the impact of microarray testing on clinical decision making. Expert Opin. Pharmacother. 2006, 7 (15), 2069-78. (5) Gant, T. W. Classifying toxicity and pathology by gene-expression profile-taking a lead from studies in neoplasia. Trends Pharmacol. Sci. 2002, 23 (8), 388-93. (6) Kim, S.; Dougherty, E. R.; Shmulevich, I.; Hess, K. R.; Hamilton, S. R.; Trent, J. M.; Fuller, G. N.; Zhang, W. Identification of combination gene sets for glioma classification. Mol. Cancer Ther. 2002, 1 (13), 1229-36. (7) Verrills, N. M. Clinical proteomics: present and future prospects. Clin. Biochem. Rev. 2006, 27 (2), 99-116. (8) Templin, M. F.; Stoll, D.; Schwenk, J. M.; Potz, O.; Kramer, S.; Joos, T. O.; Protein microarrays: promising tools for proteomic research. Proteomics 2003, 3 (11), 2155-66. (9) Paweletz, C. P.; Charboneau, L.; Bichsel, V. E.; Simone, N. L.; Chen, T.; Gillespie, J. W.; Emmert-Buck, M. R.; Roth, M. J.; Petricoin, I. E.; Liotta, L. A. Reverse phase protein microarrays which capture disease progression show activation of pro-survival pathways at the cancer invasion front. Oncogene 2001, 20 (16), 1981-9. (10) Nishizuka, S.; Charboneau, L.; Young, L.; Major, S.; Reinhold, W. C.; Waltham, M.; Kouros-Mehr, H.; Bussey, K. J.; Lee, J. K.; Espina, V.; Munson, P. J.; Petricoin, E. 3rd; Liotta, L. A.; Weinstein, J. N. Proteomic profiling of the NCI-60 cancer cell lines using new high-density reverse-phase lysate microarrays. Proc. Natl. Acad. Sci. U.S.A. 2003, 100 (24), 14229-34. (11) Espina, V.; Mehta, A. I.; Winters, M. E.; Calvert, V.; Wulfkuhle, J.; Petricoin, E. F., 3rd; Liotta, L. A. Protein microarrays: molecular profiling technologies for clinical specimens. Proteomics 2003, 3 (11), 2091-100. (12) Melton, L. Protein arrays: proteomics in multiplex. Nature 2004, 429 (6987), 101-7. (13) Wulfkuhle, J. D.; Paweletz, C. P.; Steeg, P. S.; Petricoin, E. F., III; Liotta, L. Proteomic approaches to the diagnosis, treatment, and monitoring of cancer. Adv. Exp. Med. Biol. 2003, 532 59-68. (14) Grubb, R. L.; Calvert, V. S.; Wulkuhle, J. D.; Paweletz, C. P.; Linehan, W. M.; Phillips, J. L.; Chuaqui, R.; Valasco, A.; Gillespie, J.; Emmert-Buck, M.; Liotta, L. A.; Petricoin, E. F. Signal pathway profiling of prostate cancer using reverse phase protein arrays. Proteomics 2003, 3 (11), 2142-6. (15) Akkiprik, M.; Nicorici, D.; Cogdell, D.; Jia, Y. J.; Hategan, A.; Tabus, I.; Yli-Harja, O.; Y, D.; Sahin, A.; Zhang, W. Dissection of signaling pathways in fourteen breast cancer cell lines using reverse-phase protein lysate microarray. Technol. Cancer Res. Treat. 2006, 5 (6), 543-52. (16) Jiang, R.; Mircean, C.; Shmulevich, I.; Cogdell, D.; Jia, Y.; Tabus, I.; Aldape, K.; Sawaya, R.; Bruner, J. M.; Fuller, G. N.; Zhang, W. Pathway alterations during glioma progression revealed by reverse phase protein lysate arrays. Proteomics 2006, 6 (10), 296471.

research articles

Signaling Pathways in Cancer (17) Mircean, C.; Shmulevich, I.; Cogdell, D.; Choi, W.; Jia, Y.; Tabus, I.; Hamilton, S. R.; Zhang, W. Robust estimation of protein expression ratios with lysate microarray technology. Bioinformatics 2005, 21 (9), 1935-42. (18) Tabus, I.; Hategan, A.; Mircean, C.; Rissanen, J.; Shmulevich, I.; Zhang, W.; Astola, J. Nonlinear modeling of protein expressions in protein arrays. IEEE Trans. Signal Process. 2006, 54 2394-2407. (19) Bolstad, B. M.; Irizarry, R. A.; Astrand, M.; Speed, T. P. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 2003, 19 (2), 185-93. (20) Dudoit, S F. J.; and Speed T. P. Comparison of discrimination methods for the classification of tumors using gene expression data. J. Am. Stat. Assoc. 2002, 97 77-87. (21) Benjamini, Y. a. H. Y. Controlling the False Discovery Rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. 1995, 57B, 289-300. (22) Benjamini, Y.; Drai, D.; Elmer, G.; Kafkafi, N.; Golani, I. Controlling the false discovery rate in behavior genetics research. Behav. Brain Res. 2001, 125 (1-2), 279-84. (23) Storey, J. D.; Tibshirani, R. Statistical significance for genomewide studies. Proc. Natl. Acad. Sci. U.S.A. 2003, 100 (16), 9440-5. (24) Suzuki, R.; Shimodaira, H. Pvclust: an R package for assessing the uncertainty in hierarchical clustering. Bioinformatics 2006, 22 (12), 1540-2. (25) Cullinane, C.; Dorow, D. S.; Kansara, M.; Conus, N.; Binns, D.; Hicks, R. J.; Ashman, L. K.; McArthur, G. A.; Thomas, D. M. An in vivo tumor model exploiting metabolic response as a biomarker for targeted drug development. Cancer Res. 2005, 65 (21), 96336. (26) Soussi, T.; Ishioka, C.; Claustres, M.; Beroud, C. Locus-specific mutation databases: pitfalls and good practice based on the p53 experience. Nat. Rev. Cancer 2006, 6 (1), 83-90. (27) Ji, H.; Sharpless, N. E.; Wong, K. K. EGFR targeted therapy: view from biological standpoint. Cell Cycle 2006, 5 (18), 2072-6. (28) Kamath, S.; Buolamwini, J. K. Targeting EGFR and HER-2 receptor tyrosine kinases for cancer drug discovery and development. Med. Res. Rev. 2006, 26 (5), 569-94. (29) Ciardiello, F.; De Vita, F. Epidermal growth factor receptor (EGFR) inhibitors in cancer therapy. Prog. Drug Res. 2005, 63, 93-114. (30) Pal, S. K.; Pegram, M. Epidermal growth factor receptor and signal transduction: potential targets for anti-cancer therapy. Anticancer Drugs 2005, 16 (5), 483-94. (31) Rozengurt, E.; Guha, S.; Sinnett-Smith, J. Gastrointestinal peptide signalling in health and disease. Eur. J. Surg. Suppl. 2002, (587), 23-38. (32) Kleespies, A.; Jauch, K. W.; Bruns, C. J. Tyrosine kinase inhibitors and gemcitabine: new treatment options in pancreatic cancer? Drug Resist. Updates 2006, 9 (1-2), 1-18. (33) Kain, K. H.; Gooch, S.; Klemke, R. L. Cytoplasmic c-Abl provides a molecular ‘Rheostat’ controlling carcinoma cell survival and invasion. Oncogene 2003, 22 (38), 6071-80. (34) Jones, A. V.; Cross, N. C. Oncogenic derivatives of platelet-derived growth factor receptors. Cell. Mol. Life Sci. 2004, 61 (23), 291223. (35) Dibb, N. J.; Dilworth, S. M.; Mol, C. D. Switching on kinases: oncogenic activation of BRAF and the PDGFR family. Nat. Rev. Cancer 2004, 4 (9), 718-27. (36) McCubrey, J. A.; Steelman, L. S.; Chappell, W. H.; Abrams, S. L.; Wong, E. W.; Chang, F.; Lehmann, B.; Terrian, D. M.; Milella, M.; Tafuri, A.; Stivala, F.; Libra, M.; Basecke, J.; Evangelisti, C.; Martelli, A. M.; Franklin, R. A. Roles of the Raf/MEK/ERK pathway in cell growth, malignant transformation and drug resistance. Biochim. Biophys. Acta 2006, in press. (37) Mitsuuchi, Y.; Johnson, S. W.; Selvakumaran, M.; Williams, S. J.; Hamilton, T. C.; Testa, J. R. The phosphatidylinositol 3-kinase/ AKT signal transduction pathway plays a critical role in the expression of p21WAF1/CIP1/SDI1 induced by cisplatin and paclitaxel. Cancer Res. 2000, 60 (19), 5390-4.

(38) Cho, G. W.; Shin, S. M.; Namkoong, H.; Kim, H. K.; Ha, S. A.; Hur, S. Y.; Kim, T. E.; Chai, Y. G.; Kim, J. W. The phosphatidylinositol 3-kinase/Akt pathway regulates the HCCR-1 oncogene expression. Gene 2006, 384 18-26. (39) Mungamuri, S. K.; Yang, X.; Thor, A. D.; Somasundaram, K. Survival signaling by Notch1: mammalian target of rapamycin (mTOR)-dependent inhibition of p53. Cancer Res. 2006, 66 (9), 4715-24. (40) Hardie, D. G.; Carling, D.; Carlson, M. The AMP-activated/SNF1 protein kinase subfamily: metabolic sensors of the eukaryotic cell? Annu. Rev. Biochem. 1998, 67, 821-55. (41) Rattan, R.; Giri, S.; Singh, A. K.; Singh, I. 5-Aminoimidazole-4carboxamide-1-beta-D-ribofuranoside inhibits cancer cell proliferation in vitro and in vivo via AMP-activated protein kinase. J. Biol. Chem. 2005, 280 (47), 39582-93. (42) Motoshima, H.; Goldstein, B. J.; Igata, M.; Araki, E. AMPK and cell proliferation-AMPK as a therapeutic target for atherosclerosis and cancer. J. Physiol. 2006, 574, (Pt 1), 63-71. (43) Levav-Cohen, Y.; Haupt, S.; Haupt, Y. Mdm2 in growth signaling and cancer. Growth Factors 2005, 23 (3), 183-92. (44) Harfouche, R.; Hassessian, H. M.; Guo, Y.; Faivre, V.; Srikant, C. B.; Yancopoulos, G. D.; Hussain, S. N. Mechanisms which mediate the antiapoptotic effects of angiopoietin-1 on endothelial cells. Microvasc. Res. 2002, 64 (1), 135-47. (45) Pervaiz, S. Resveratrol-from the bottle to the bedside? Leuk. Lymphoma 2001, 40 (5-6), 491-8. (46) Kim, R.; Emi, M.; Tanabe, K.; Toge, T. Therapeutic potential of antisense Bcl-2 as a chemosensitizer for cancer therapy. Cancer 2004, 101 (11), 2491-502. (47) Shanafelt, T. D.; Kay, N. E. The clinical and biologic importance of neovascularization and angiogenic signaling pathways in chronic lymphocytic leukemia. Semin. Oncol. 2006, 33 (2), 17485. (48) Farahani, M.; Treweeke, A. T.; Toh, C. H.; Till, K. J.; Harris, R. J.; Cawley, J. C.; Zuzel, M.; Chen, H. Autocrine VEGF mediates the antiapoptotic effect of CD154 on CLL cells. Leukemia 2005, 19 (4), 524-30. (49) Messmer, D.; Kipps, T. J. CD154 gene therapy for human B-cell malignancies. Ann. N. Y. Acad. Sci. 2005, 1062, 51-60. (50) Kopnin, B. Genetic events responsible for colorectal tumorigenesis: achievements and challenges. Tumori 2003, 79 (4), 23543. (51) Yu, B. W.; Nguyen, D.; Anderson, S.; Allegra, C. A. Phosphorothioated antisense c-myc oligonucleotide inhibits the growth of human colon carcinoma cells. Anticancer Res. 1997, 17 (6D), 4407-13. (52) Johnson, D. G.; Degregori, J. Putting the Oncogenic and Tumor Suppressive Activities of E2F into Context. Curr. Mol. Med. 2006, 6 (7), 731-8. (53) Moore, M. G, W. L.; Francis, M. J.; Peehl, D. M.; Cohen, P. Novel stimulatory role for insulin-like growth factor binding protein-2 in prostate cancer cells. Int. J. Cancer 2003, 105 (1), 14-19. (54) Calvo, K. R.; Liotta, L. A.; Petricoin, E. F. Clinical proteomics: from biomarker discovery and cell signaling profiles to individualized personal therapy. Biosci. Rep. 2005, 25 (1-2), 107-25. (55) Araujo, R. P.; Petricoin, E. F.; Liotta, L. A. A mathematical model of combination therapy using the EGFR signaling network. Biosystems 2005, 80 (1), 57-69. (56) Gulmann, C.; Sheehan, K. M.; Kay, E. W.; Liotta, L. A.; Petricoin, E. F. 3rd, Array-based proteomics: mapping of protein circuitries for diagnostics, prognostics, and therapy guidance in cancer. J. Pathol. 2006, 208 (5), 595-606. (57) Faivre, S.; Djelloul, S.; Raymond, E. New paradigms in anticancer therapy: targeting multiple signaling pathways with kinase inhibitors. Semin. Oncol. 2006, 33 (4), 407-20.

PR070184H

Journal of Proteome Research • Vol. 6, No. 7, 2007 2767

Analysis of Signaling Pathways in 90 Cancer Cell Lines by Protein

Recommend Documents